Understanding and Evaluating HeatRisk Forecasts for Public Health: A Comprehensive Assessment of Heat-Related Impacts Using Syndromic Surveillance Information

Heat prevention and mitigation efforts, including communication and outreach, rely on real-time monitoring of weather conditions. In addition, public health practitioners and emergency planners benefit from heat forecast products and tools that are accurate, easy-to-interpret, and offer sufficient lead time to implement necessary interventions. Toward this end, the experimental HeatRisk Forecast Tool was developed by National Oceanic and Atmospheric Administration’s (NOAA) National Weather Service (NWS) and the Centers for Disease Control and Prevention (CDC). HeatRisk is a multi-tiered framework that applies a consistent approach, incorporating high-resolution climatology, locally relevant heat-mortality relationships, and peer-reviewed scientific methods, into a numeric and color-based seven-day forecast system. Specifically, the tool uses an impact-based framework to indicate how risky the prevailing levels of heat are in a specific area and delineates exposure risk into a five-level scale: Green (0)—No Risk; Yellow (1)—Minor Risk; Orange (2)—Moderate Risk; Red (3)—Major Risk; Magenta (4)—Extreme Risk.

For this assessment, an evaluation of HeatRisk tool was conducted using emergency department (ED) visits from CDC’s National Syndromic Surveillance Program (NSSP). County-level data on ED visits occurring during summer months (May–September) of 2018–2023 were extracted from NSSP’s Electronic Surveillance System for the Early Notification of Community-Based Epidemics. Daily numbers of heat-related illness (HRI) and all-cause ED visits were tabulated for the following 12 states, located in different regions of the Continental United States: Midwest (Michigan, Wisconsin), Northeast (Maine, Vermont, New York), Southeast (Georgia, North Carolina, Tennessee, Virginia), and West (Colorado, New Mexico, Washington). The primary exposure measure was the population-weighted average daily, county-level HeatRisk levels obtained from weather stations. A conditional quasi-Poisson regression with strata by county, year, month, and day of the week was used to estimate the relative risks (RRs) and 95% confidence interval (CI) of ED visits associated with HeatRisk levels. Furthermore, the modelling approach controlled for seasonality, holidays, and estimated overall cumulative RRs [CI] for HeatRisk categories over a 2-day lag period.

Daily HRI rates and HeatRisk levels were positively correlated with state-specific Spearman’s rank-order correlation coefficients ranging between 0.56 and 0.85 (Median =0.78). During the summer months, any day with HeatRisk levels at or above minor risk (i.e., Yellow, Orange, Red, or Magenta) was associated with a significant increase in risk of HRI ED visits (RR: 6.301; 95% CI: 5.972–6.652); the association between HeatRisk levels and total all-cause ED visits was significant but the magnitude of risk was smaller (RR: 1.031; 95% CI: 1.029–1.033). Of note, regional differences in RRs were observed and individual factors such as age and sex influenced relationships between ED visits and HeatRisk levels.

Heat exposure continues to be a public health concern and with climate change, heat events are getting longer, more frequent, and more intense. Robust tools for forecasting heat exposure with sufficient lead time and timely mechanisms for tracking and reporting health effects, especially during extreme heat emergencies, can strengthen public health-centric efforts to prevent heat-related health harms.